Auto-suggest helps you quickly narrow down your search results by suggesting possible matches as you type.

- JMP User Community
- :
- Discussions
- :
- Design comparison

- Subscribe to RSS Feed
- Mark Topic as New
- Mark Topic as Read
- Float this Topic for Current User
- Bookmark
- Subscribe
- Printer Friendly Page

Highlighted

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Jun 4, 2019 1:09 PM
(3305 views)

I am comparing 4 different designs with different runs and trying to figure out which one is better and want to keep number of runs minimum. Attached is the design comparison I did. If i look at the power, 9run design has lowest power and rest of the runs are better options but if I look at the correlations 9 run and 18 run looks good and 12 run and 15 run has correlation issues. Fraction of design space plot is also good for 18 run design. Is there any other thing that can help with the decision that I am missing. Also if I look at the absolute correlation table, average correlation for all the designs is below 0.3 and there is no confounding. Is it safe to say that 12 and 15 run design are good although the color map shows some correlations (light blue/grey color)?

1 ACCEPTED SOLUTION

Accepted Solutions

Highlighted

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

I would base your decision in this case on the power information. If you really are after detecting effects that are the same magnitude as random error (RMSE) then the only design I would personally consider is the 18 run design. This is what you have specified by defaulting the effect values to 1 and the RMSE to 1. The power is too low for me in the other designs you are considering. The power is the probability of detecting an effect of the size specified if the effect really occurs.

If the model effects you are interested in detecting are somewhat larger than the RMSE, then try specifying higher values for the model effects and see what impact this has on the power and other design comparison aspects.

2 REPLIES 2

Highlighted

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

I would base your decision in this case on the power information. If you really are after detecting effects that are the same magnitude as random error (RMSE) then the only design I would personally consider is the 18 run design. This is what you have specified by defaulting the effect values to 1 and the RMSE to 1. The power is too low for me in the other designs you are considering. The power is the probability of detecting an effect of the size specified if the effect really occurs.

If the model effects you are interested in detecting are somewhat larger than the RMSE, then try specifying higher values for the model effects and see what impact this has on the power and other design comparison aspects.

Highlighted
##

- Mark as New
- Bookmark
- Subscribe
- Mute
- Subscribe to RSS Feed
- Get Direct Link
- Email to a Friend
- Report Inappropriate Content

Re: Design comparison

@malcolm_moore1 thank you for your response. I selected 18 run design for the same reason.

Article Labels

There are no labels assigned to this post.